The influence of visual vs verbal on verbal depiction on assortment processing perceived variety and choice overload

Journal of Consumer Research, Inc.

The “Visual Preference Heuristic”: The Influence of Visual versus Verbal Depiction onAssortment Processing, Perceived Variety, and Choice OverloadAuthor(s): Claudia Townsend and Barbara E. KahnSource: Journal of Consumer Research, (-Not available-), p. 000Published by: The University of Chicago PressStable URL: http://www.jstor.org/stable/10.1086/673521 .

Accessed: 16/10/2013 16:53

Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at .http://www.jstor.org/page/info/about/policies/terms.jsp

.JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range ofcontent in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new formsof scholarship. For more information about JSTOR, please contact [email protected].

.

The University of Chicago Press and Journal of Consumer Research, Inc. are collaborating with JSTOR todigitize, preserve and extend access to Journal of Consumer Research.

http://www.jstor.org

This content downloaded from 75.26.233.69 on Wed, 16 Oct 2013 16:53:06 PMAll use subject to JSTOR Terms and Conditions

http://www.jstor.org/action/showPublisher?publisherCode=ucpress

http://www.jstor.org/stable/10.1086/673521?origin=JSTOR-pdf

http://www.jstor.org/page/info/about/policies/terms.jsp


000

� 2013 by JOURNAL OF CONSUMER RESEARCH, Inc. ● Vol. 40 ● February 2014All rights reserved. 0093-5301/2014/4005-0010$10.00. DOI: 10.1086/673521

Please use DOI when citing. Page numbers are not final.

The “Visual Preference Heuristic”: TheInfluence of Visual versus Verbal Depictionon Assortment Processing, PerceivedVariety, and Choice Overload

CLAUDIA TOWNSENDBARBARA E. KAHN

The “visual preference heuristic” suggests that consumers prefer visual to verbaldepiction of information in a product assortment. Images produce greater percep-tions of variety than text, which is appealing in assortment selection, but can resultin choice complexity and overload when choice sets are large and preferencesare unknown, suggesting a moderator for Iyengar and Lepper’s results. Eye-track-ing results reveal that the natural gestalt processing of individual visual stimuli, ascompared to the piecemeal processing of individual textual stimuli, affects theprocessing of the assortment as a whole with visual (compared to verbal) pre-sentation facilitating a faster, though more haphazard, scanning of the assortment.While the less systematic processing that results from visual presentation feelseasier, it is not ideal for larger assortments resulting in higher complexity ratingsand choice overload than with text depiction. These findings reveal that, like manyheuristics, preference for visual depiction may be overapplied.

As online access becomes increasingly common, con-sumers are becoming multichannel shoppers, as likely

to view product assortments online as offline. The onlineenvironment allows the number of options offered to bequite large as there are not the constraints that exist in brickand mortar environments. Instinctively, consumers prefermore variety, yielding a greater chance that their preferenceswill be matched (Kahn and Lehmann 1991). Yet we alsoknow there is a downside to too much variety; it can beoverwhelming and lead to choice overload. Product assort-ments that are large and complex can cause consumers to

Claudia Townsend ([email protected]) is assistant professor ofmarketing, University of Miami, 5250 University Drive, 501 Kosar/EpsteinBuilding, Coral Gables, FL 33124. Barbara E. Kahn ([email protected]) is Patty and Jay H. Baker Professor of Marketing and Directorof the Jay H. Baker Retailing Center, The Wharton School, University ofPennsylvania, 763 John M. Hunstman Hall, 3730 Walnut Street, Phila-delphia, PA 19104. The authors thank Laura Peracchio, Rebecca Hamilton,and the reviewers for their helpful input. The authors also thank ChelseaWise for help with collecting the eye tracking data.

Laura Peracchio served as editor and Rebecca Hamilton served as as-sociate editor for this article.

Electronically published October 16, 2013

feel overwhelmed and dissatisfied or to opt not to make achoice at all (Huffman and Kahn 1998; Iyengar and Lepper2000; Jacoby, Speller, and Berning 1974). Also, consumersdissatisfied with the shopping process because of large prod-uct assortment are less likely to revisit the store or repur-chase (Fitzsimons, Greenleaf, and Lehmann 1997). So whilegreater variety is generally preferred when choosing amongassortments, when consumers are faced with actual optionselection, variety can make the process more difficult andcomplex.

The actual variety of an assortment, as in the number ofstock-keeping units (SKUs) within a category, is usuallyreferred to in the retailing industry as the assortment depth(Mantrala et al. 2009). Further, perceptions of variety candiffer even when actual variety is held constant. In fact, thelikelihood that consumers will purchase from a retail site ispositively related to the perceived variety that they thinkthe assortment offers, rather than the actual variety (Bro-niarczyk, Hoyer, and McAlister 1998). Huffman and Kahn(1998) showed that when actual variety in choice sets waslarge and overwhelming, perceived variety and hence con-sumer satisfaction could be controlled through both the wayinformation about the options was presented and consumerinput in the product examination process. Others found that


mailto:[email protected]




000 JOURNAL OF CONSUMER RESEARCH


perceived variety varied as a function of the organizationof the assortment (Hoch, Bradlow, and Wansink 1999; Kahnand Wansink 2004), the relative symmetry in the frequenciesof the items in the assortments (Young and Wasserman2001), and individual differences (Morales et al. 2005).

Here we consider how perceived variety of an assortmentonline varies as a function of whether the options are pre-sented using words or images. While assortments can bepresented using a combination of the two, in many casesone information mode dominates in initial scanning, partic-ularly when viewing a retailer’s offering on a small screen.For example, many applications for shopping on a tablet orsmartphone (e.g., those from Zappos, Gilt, or Amazon) pri-oritize one mode on the initial screen (e.g., visual) and thenthe other mode on subsequent screens (e.g., more detailedverbal description).

Holding constant the number of options, assortment or-ganization, and amount of information offered about eachoption, we show how presentation format, whether presentedvisually or verbally, influences perceptions of variety andcomplexity. We find that when selecting an assortment ofoptions, greater perceived variety is preferred. However, inthe choice stage, when considering individual options, toomuch perceived variety can increase complexity and causeconsumers to delay or opt out of choice. We predict thatvisual depiction, which increases perceived variety, is gen-erally preferred to verbal depiction in the first stage, theassortment decision. However, subsequently, when individ-ual options are considered, visual depiction is not alwaysoptimal as, with larger choice sets, it can increase perceivedcomplexity and thus lead to greater choice deferral. We focuson conditions when preferences are unknown, in line withmost of the choice overload literature (Scheibehenne, Grei-feneder, and Todd 2010).

Past research examining the differences in processing ofimages versus words provides an explanation for these find-ings. Images are processed in a gestalt manner that is faster,less deliberate, and thus feels easier than the piecemeal pro-cessing that is associated with verbal stimuli. Additionally,since gestalt processing allows for parallel rather than se-quential recognition of attributes, it is easier to identify at-tribute interactions (Holbrook and Moore 1981; Veryzer andHutchinson 1998), which increases the perceived variety ofthe assortment. Together these aspects of gestalt processingresult in a preference for visually presented assortmentchoice sets. Thus we predict the existence of a “visual pref-erence heuristic,” whereby consumers generally opt for vi-sual presentation over verbal.

However, in large choice sets, the increased perceptionsof variety are accompanied by increased perceptions of com-plexity, which can result in choice overload. When consum-ers are forced to examine many options for choice, the faster,easier, and less deliberate gestalt processing that naturallyoccurs with images is less desirable (Hoch and Deighton1989; Meyer 1986, 1987). In addition, we propose that themode of processing used to examine individual images car-ries over to the manner in which the entire assortment is

considered, resulting in a faster and less systematic approachfor visual as opposed to verbal assortments. Compared togestalt processing, the slower, more deliberate piecemealprocessing of words, while making verbal assortments lesspreferred, is a more effective form of processing of largechoice sets both for the individual items as well as the fullassortment. It follows in cases when preferences are notknown and further scrutiny of the large choice set is re-quired, the preference for visual depiction may not alwaysbe optimal. As such this research offers three key contri-butions:

1. We identify and find evidence of a “visual prefer-ence heuristic”: consumers choose visual over ver-bal depiction of choice sets, regardless of choice setsize.

2. We observe the influence of presentation format onchoice behavior revealing visual presentation to besuboptimal for large choice sets when preferencesare unknown.

3. We show that the manner in which individual op-tions are processed has a carryover effect to theexamination of the entire assortment.

THEORETICAL BACKGROUND ANDHYPOTHESIS DEVELOPMENT

Distinctive Processing of Images and Words

Studies using brain-imaging techniques show that, whilethere is a common neural network activated by both verbaland visual processing, there are also specific areas of brainactivation for each form: the left inferior parietal lobule forthe processing of words versus the right middle occipitalgyrus for processing pictures. In considering higher levelprocessing such as memory, expectation, and attention, re-searchers examine event-related potentials (ERPs), neuralsigns that reflect brain activity. Studies comparing ERPs ofwords and pictures find consistent differences between thetwo (see Khateb et al. [2002] for review). Paivio’s dualcoding theory (1971, 1986) describes two distinct but in-terconnected systems—one visual and one verbal.

In addition to these differences in brain activation, imagesare also processed more quickly and automatically, and theconnection between an image and its meaning is more directthan it is for words (Luna and Peracchio 2003; Paivio 1971;Unnava and Burnkrant 1991). Carr et al. (1982) investigatedthe use of primes and found that images offer faster andmore automatic access to their meanings than to their names,while words have faster and more automatic access to theirnames than their meanings. Images also trigger more emo-tional processing than do words (Hsee and Rottenstreich2004; Lee, Amir, and Ariely 2009; Lieberman et al. 2002).

Additional evidence for the automatic and even uncon-scious nature of image processing can be found by observinghow aspects of aesthetics, such as symmetry or certain pro-portions (e.g., the golden ratio), can influence perceptionsof attractiveness without explicit awareness (Arnheim 1974;



TOWNSEND AND KAHN 000


Birkhoff 1933). Studies using images of facial expressionsshow that the meanings of facial expressions are automat-ically processed (e.g., Stenberg, Wiking, and Dahl 1998).Because it is faster and more automatic, the processing ofimages rather than words likely feels easier.

Gestalt versus Piecemeal Processing. In addition to thesemore automatic reactions, there are also differences in themanner in which information is interpreted. Words and num-bers must be processed sequentially or in a piecemeal man-ner, whereas an image can be processed all at once (Hart1997). Similarly, Paivio (1986) describes the imagery sys-tem as able to work in parallel, while the verbal systemworks sequentially. However, Sharps and Nunes (2002) sug-gest that it is inappropriate to think of these two modes ofdepiction as two rigidly separated processes operating inisolation, but that it is better to think of these processingstyles as lining up along a continuum. While they show thatconsistent with the prior thinking, images tend to be pro-cessed in a more gestalt fashion and verbal material in amore piecemeal fashion, they find that within the visualmodality alone, one can illustrate stimuli in a piecemeal aswell as a gestalt manner; however, the latter is more realisticand prevalent. The visual stimuli we use in this researchwill be of the more common format that can be processedin a gestalt manner.

Preference for Choice Sets. Because of the natural gestaltprocessing that occurs with images versus the slower andmore deliberate piecemeal processing that occurs withwords, an assortment presented in visual format seems fasterand easier to process (Holbrook and Moore 1981; Veryzerand Hutchinson 1998). We predict that consumers are likelyto choose the assortment that they perceive will allow themto engage in a fast and easy decision process and thereforehypothesize the following:

H1: Consumers will generally prefer visual over verbaldepictions of choice options.

We refer to this as the visual preference heuristic and nextconsider a boundary condition for this prediction. If thispreference is driven, at least in part, by the perception thatviewing images allows for an easier choice process, then insituations when this seems not to be the case, we wouldexpect the preference to go away. While our prediction ofthe visual preference heuristic is based on the processing ofimages versus words, consumers likely only recognize ageneral choice process and not individual parts (processing,evaluation, etc.). Therefore, making the choice process seemmore difficult with images than with words ought to atten-uate the visual preference. We test this by manipulating howeasy it is to discern differences in attribute levels in visualversus verbal presentation. We predict the following:

H2: When visual attribute discernment is more diffi-cult and therefore choice from visual presentationseems more difficult than from verbal presenta-tion, the preference for visual depiction will beattenuated.

To determine whether preference for visual presentation isoptimal, we consider how assortment presentation mightinfluence assortment processing, the resulting assortmentperceptions, and subsequently choice behavior.

Processing of the Assortment as a Whole. We predict thatthe differences in processing of images versus words influ-ence, not only the processing of the individual items withinthe choice sets, but also how the choice sets are examinedas a whole. We proffer that the piecemeal processing re-quired for each verbal stimulus motivates a more systematicprocessing of the choice options as a set, yielding a moremethodical evaluation, more time spent, and a lower like-lihood of skipping or bypassing information about individualchoice options. When options are presented verbally andconsumers process in a piecemeal fashion, they read theinformation about each option in a more structured system-atic manner—sequentially horizontally or vertically. In con-trast, the gestalt processing that naturally occurs with eachvisual stimulus likely encourages a faster, less systematicprocessing of a visual choice set where options are evaluatedmore randomly.

The result is that with gestalt processing, less time is spentconsidering each option, the examination path between op-tions is more haphazard, and there is a greater likelihoodthat options may be overlooked. However, the impact of aless systematic style of processing on such factors shoulddepend on the number of options in the choice set. Forexample, the likelihood that any option will be skipped in-creases as the number of options in the set increases. Sim-ilarly, with larger choice sets there are more degrees offreedom in terms of movement paths. We therefore hypoth-esize the following:

H3: In larger choice sets, visually depicted assortmentswill be processed in a less systematic manner thanverbally depicted assortments.

Additionally, we predict that this more systematic processingthat occurs with verbal rather than visual depiction in largerchoice sets—both at the individual item level as well asacross the assortment—will result in a more comprehensive,less idiosyncratic processing of the items in the set. More-over, there ought to be measurable behavioral differencesin response to the different types of processing. Specifically,the more comprehensive and less idiosyncratic processingis likely more effective, resulting in better recognition andretention of items in the assortment. However, we also ex-pect this behavioral response will depend on the size of thechoice set. In smaller choice sets, where a cursory exami-nation of options is likely adequate, we do not expect tosee differences in processing as a function of depiction.However, in larger choice sets where the more systematicprocessing that is motivated by verbal depiction would bebeneficial, we would expect to see a difference across de-piction modes. We therefore predict the following:

H4: In larger choice sets, the systematic processingthat naturally occurs with verbally depicted items





(as opposed to visually depicted items) will resultin better recognition and retention of more itemsin the assortment. In smaller choice sets, this willnot be the case.

Hypotheses 3 and 4 suggests that the preference for visualdepiction of choice sets is not always optimal and may differas the stage of choice process (and hence the consumer goal)changes. To support this intuition, we next discuss how theseprocessing differences generated by format can influenceperceptions of variety and complexity of the choice set. Webegin with an overview of the literature that shows whythose assortment perceptions matter.

Variations on variety and complexity perceptions by stageof choice.—To attract consumers to an assortment, managerstry to increase perceptions of variety, the positive and at-tractive aspect of assortment, while minimizing perceptionsof complexity, the negative aspect of assortment, that canoccur during the choice stage and cause choice overload.Scheibehenne et al. (2010) provided a meta-analysis of theresearch investigating choice overload. This meta-analysisfound a mean effect size of zero for choice overload butfound considerable variance between studies. As Chernev,Bockenholt, and Goodman (2010) suggest, the interestingquestion is not whether choice overload occurs but when itoccurs, and thus the key is to identify the conditions thatcan minimize or maximize choice overload. The meta-anal-ysis argues that one necessary precondition for overload isa lack of familiarity with, or prior preferences for, items inthe choice set. Another precondition is that there is no dom-inant option in the set. In our studies we only examine choicesets where preferences are unknown and there are no dom-inant options.

Existing research offers some prescriptions for maintain-ing the same amount of actual variety but mitigating thenegative effects of perceived complexity. For example, Hochet al. (1999), Kahn and Wansink (2004), and Young andWasserman (2001) suggest the organization of the optionscan be varied so that the same assortment can be offered ina manner that minimizes perceived complexity. More re-cently, Mogilner, Rudnick, and Iyengar (2008) demonstratedthe “mere categorization effect,” that the number and contentof category labels can affect perceptions of variety at theassortment stage and customer satisfaction at the choicestage. This complements the work by Morales et al. (2005),which showed that matching external structure in the storewith internal personal categorization of the category can alsominimize the complexity of large assortments. All of thissuggests that while perceptions of variety and complexitymay be highly correlated, there are ways in which the twocan be disentangled. Specifically, perceived variety is rel-evant when consumers are at the first stage of the choiceprocess, and their goal is to choose an assortment. Perceivedcomplexity is relevant when consumers are at the secondstage of the choice process and their goal is to choose anoption from within the assortment. We propose that visual/verbal presentation format is another important moderator

for understanding how perceptions of variety and complex-ity may vary at different stages of the choice process.

Variety perceptions.—With respect to perceived variety,we predict a main effect. We predict that the natural gestaltprocessing of images versus piecemeal processing of wordswill influence perceptions of variety of the assortment. Whenchoice options are described in words, each attribute of theoption is processed individually, making consideration ofthe interactions among the various attributes difficult. Incontrast, when choice options are presented in images, allattributes are processed simultaneously. The recognition ofattribute interactions as well as main effects increases thenumber of permutations, and thus perceptions of varietyincrease. Hence, we hypothesize the following:

H5: Perceived variety will be higher for visually ver-sus verbally depicted choice sets.

Complexity perceptions.—Perceived complexity occurswhen the consumer goal is to choose among options withinthe assortment, and at this stage the size of the choice setbecomes relevant. When choice sets are at least moderatelycomplex and unfamiliar, processing in a gestalt or alterna-tive-based manner is difficult (Hoch and Deighton 1989;Meyer 1986, 1987). When information can be processed insmaller chunks, for example, in a piecemeal fashion, theinformation is easier for nonexperts to process (Alba andHutchinson 1987; Chase and Simon 1973). Huffman andKahn (1998) found that for large choice sets, presentinginformation in an attribute-based setting, rather than in analternative-based setting (which requires more gestalt pro-cessing) reduced perceived complexity. It follows then thatwhen considering a large choice set with unfamiliar stimuli,there may be advantages to verbal presentation over visual.Per hypothesis 3, when options are presented verbally andconsumers process in a piecemeal fashion, they consider theinformation both within and across options in a structuredsystematic manner. Taken together, we therefore predict thefollowing:

H6: In larger choice sets, respondents who see optionsdescribed verbally will perceive less complexityin the choice options than respondents who seeoptions described visually.

There is evidence to suggest that the relative importanceand preference for variety and complexity varies by stageof the choice process as consumers’ goals change. For ex-ample, Novemsky and Ratner (2003) showed that consumersbelieve in the positive effect of contrast (which can be ob-tained by adding more variety), in anticipation of con-sumption, but during the actual consumption there are nopositive advantages of the variety or contrast experienced.Construal theory (Trope and Liberman 2010) also predictsthat psychological distance from the choice alters prefer-ences. For example, consumers prefer desirable over feasiblealternatives in thinking about the future and prefer morefeasible alternatives when thinking about the present.

These two theories suggest that at the first stage, before





any actual choices are made, consumers likely prefer as-sortments with more perceived variety. Therefore if, as wepredict, visual presentation increases perception of variety(hypothesis 5), this may further reinforce hypothesis 1 andthe preference for visual depiction. However, as construaltheory predicts, when the consumer then gets closer to theactual choice decision, the concrete trade-offs of each itembecome salient and perceived complexity is more relevant.Here feasibility becomes more important than desirability(Hamilton and Thompson 2007). This is consistent withScheibehenne et al. (2010), who conclude that when con-sumers are focusing on the assortment, attractiveness tendsto correlate positively with size of the assortment; however,when consumers are focusing on determining the appeal ofa specific option, choice overload comes into play and con-sumers are overwhelmed with large choice sets. In order tomake this final choice from the assortment, consumers haveto process not just the entire assortment but also the indi-vidual options, and therefore the combination of the twoinfluences perceptions of complexity. These complexity per-ceptions then influence what will ultimately be chosen orwhether the consumer will opt out of making a choice.

It follows then when making a choice from a large choiceset with unfamiliar stimuli, there may be advantages to ver-bal presentation over visual. The more systematic (hypoth-esis 3) and comprehensive (hypothesis 4) processing thatoccurs with verbal depiction likely makes the choice set lessoverwhelming, decreasing complexity (hypothesis 6) andthus decreasing choice overload. Support for this hypothesisexists in other empirical evaluations of choice. Bawa et al.(1989) found that consumers were more promotion-sensitive(which requires a systematic evaluation of numeric infor-mation) when presented with larger assortments than withsmaller assortments. Thus, we hypothesize the following:

H7A: In a larger choice set, respondents who see op-tions described verbally will be less likely to optout of making a choice than respondents whosee options described visually. In a smallerchoice set, presentation format will not influencepropensity to opt out of making a choice.

H7B: The perception of complexity will mediate theinfluence of presentation format on choice over-load (propensity to opt out of making a choice).

So we are proposing perceptions of assortment complexityas another reason for choice deferral, in addition to the moretraditional reasons already identified in the literature thatfocus on choice options themselves, such as choice difficultyresulting from the stress of trading off attributes of optionsor preference uncertainty resulting from the difficulty infinding the most attractive option among similar alternatives(Dhar 1997).

Taken together, these seven hypotheses suggest that thereis a general preference for visual presentation of assortments,which we identify as the visual preference heuristic. More-over, our theory predicts that like other choice heuristics,this visual preference is overapplied; while visual presen-

tation may be sufficient or even optimal for smaller choicesets, with larger choice sets images result in inferior pro-cessing of the entire assortment, higher perceptions of as-sortment complexity, and choice overload.

Experiments

In five studies we offer support for these hypotheses. Instudy 1 we show that there is a general preference for visualpresentation and that this preference exists regardless ofchoice set size and also in two product categories that differon key dimensions related to the efficacy of visual/verbaldepiction. In study 2, we use a product category and set ofstimuli where differentiation on a key attribute is easier withverbal depiction than visual and show that the preferencefor visual depiction decreases. In study 3, using eye-trackingtechnology we show that with images consumers use a faster,more haphazard, and less systematic manner of examiningthe options than with verbal text presentation. The next twostudies examine the consequences of this. In study 4 we con-sider performance on a matching task to examine how visualversus verbal presentation format influences processing ef-fectiveness. Then in study 5 we show that in large choicesets visual depiction is not optimal and leads to higher per-ceptions of complexity and greater likelihood to opt out ofchoice.

In all of our experiments, we are careful to ensure thatthe information content of the presentation formats is heldconstant. We present respondents with a key that describesthe product options both visually and verbally, thus insuringequivalent information for all conditions and controlling forany variation in interpretation of the meaning of the text orimages. Also, as illustrated in the common phrase “a pictureis worth a thousand words,” images often contain more in-formation than their verbal equivalent. To ensure that anydifferences we observe are not driven by additional infor-mation that might be intuited from realistic visual depictions,we use cartoon illustrations.

STUDY 1: PREFERENCE FOR VISUALPRESENTATION FORMAT

In study 1, we test hypothesis 1 that there is a preferencefor visually presented assortments with two different sizesof choice sets. Since we believe this preference is based onthe differences in processing styles of visual versus verbaldepiction, we predict that choice set size should not affectthe general preference for visual depiction.

We also use two product categories: mutual funds andcrackers. Different product categories intrinsically lendthemselves to presentation in one format or another. Forexample, at one extreme, products such as art or fashionthat derive utility from aesthetics should have an inherentadvantage with visual presentation and would be a weaktest of our hypothesis, so we do not include such a productcategory. On the other extreme are products that are muchmore suited to textual rather than image depiction, productswith attributes that offer no obvious visual depiction and





that consumers are used to seeing in text form. Mutual fundsis an example of such a product category. In the middle ofthis spectrum would be a product like crackers, which canbe presented both visually and verbally and where attributeinteractions are not of obvious importance (e.g., flavor andshape do not likely interact). If such interactions did matter,this would offer an advantage to image depiction and there-fore be a weak test of our hypothesis. We believe the resultsfrom crackers can be extrapolated to many categories, oneswhose attributes can be expressed either verbally or visually,for example, many consumer package goods products, fur-niture (e.g., sofas, tables), or clothing (e.g., sports shoes).In fact, many of the product categories that are suggestedfor consumer-based customization, like Nike sports shoes,have this characteristic. Because our hypothesis is based onthe differences in processing styles of visual versus verbaldepiction, we predict that as long as the category can beexpressed somehow visually, the preference for visual oververbal depiction should exist.

Finally, we also ask participants about their motivationfor choice set selection. Because we believe the preferencefor visual over verbal is driven by the faster, easier, andseemingly more enjoyable processing of images, we expectthese to be the motivations that participants who select im-ages will state. Additionally, because we suspect that con-sumers identify and appreciate the greater variety that seemsto come with visual depiction, we also predict a differencein application of this motive as well.

Method

Two hundred ninety participants (40% male, Mage p 33)who were recruited using Amazon’s Mechanical Turk par-ticipated in the experiment online for compensation. Re-spondents were randomly assigned to one of two conditionsthat varied by product category—crackers or mutual funds.Additionally, we randomized between subjects the order oftwo questions—preference for presentation form (visual vs.verbal) and preference for number of options (8 vs. 27).

Respondents in the cracker conditions were told to imag-ine they were shopping for a party using an online grocerystore and they had one item left on their shopping list topurchase, crackers. They were told the website offers a va-riety of crackers and then were shown three example crack-ers. This was a key that offered the information visually andverbally together (see app. A, fig. A1). The visual repre-sentation question explained that the website offers two pre-sentation forms. Respondents were asked to select whetherthey wanted the options presented as “written descriptionsof the crackers as expressed in words” or “visual images ofthe crackers as shown in pictures.” After their choice, theywere then shown three examples of cracker options in theform they chose. They were then told that the website offerstwo varieties of cracker options, either 8 crackers or 27crackers. They were then asked to select from which varietyof crackers they would like to choose. For respondents whowere first asked about the number of options they wouldlike to see, they were shown either 8 or 27 large Xs to

demonstrate their choice before being asked about depictionstyle.

For both conditions, respondents were then shown thechoice assortment they had selected: 8 or 27 options presentedvisually or verbally. Respondents were asked to rate the as-sortment on three 9-point scales measuring perceived variety(“This assortment of crackers offers a lot of variety,” “Thisassortment of crackers gives me at least one option I like,”and “How much variety do you think there is in this assort-ment?”). They were shown the assortment again and askedquestions related to making a choice. These were our mea-sures of perceived complexity (“This assortment of crackersin too complex to consider,” “It is difficult to keep track ofall the various options in this cracker assortment,” and “Thereare too many options in this assortment of crackers”). Thesemeasures of variety and complexity are adapted from Kahnand Wansink (2004). Respondents were then asked to identifytheir motives for selecting the format and set size that theychose. They were given a list of 10 motives and asked toselect up to three that best describes their reasons.

Respondents in the mutual funds conditions followed asimilar procedure. However, they were told to imagine theyhad just received a $500 bonus from their job and that theywere considering investing it in a mutual fund. They werethen given a description of three attributes on which thesemutual funds differed: fees, mean and standard deviation ofannual return, and region of investment. They were thenshown a key illustrating the various ways in which the mu-tual funds varied, and again the key displayed the infor-mation both in pictures and in words (see app. A, fig. A2).The rest of the survey was the same as for the crackers withappropriate product class edits.

All participants then filled out the 22-item visual versusverbal information processing scale (Childers, Houston, andHeckler 1985), basic demographic questions, and were askedto rate their concern with giving the correct response in thestudy on a 7-point scale.

Results

Twenty-seven participants (9.3%) who indicated that theyhad minimal concern with giving correct answers in thestudy (1 on 7-point scale) were excluded from this analysis,leaving a final sample of 263. (In future studies we excludedrespondents based on the same criteria and report the initialsample based on the final number of participants.)

Assortment Presentation Preference. The majority (64.6%,t(262) p 4.97, p ! .001) of respondents opted for optionspresented visually rather than verbally. This supports hy-pothesis 1. Looking only at respondents who were first askedabout the number of options in the choice set, this did notinfluence preference for information representation with60% (8 options) and 69% (27 options) selecting images(x2(1, N p 136) p 1.11, p p .29). Additionally, respon-dents who saw crackers were not significantly more likelyto opt for visual presentation than those who saw mutualfunds (crackers: 67.2%, mutual funds: 61.9%; x2(1, N p





FIGURE 1

STUDY 1: VARIETY AND COMPLEXITY RATINGS

NOTE.—Variety ratings (A) and complexity ratings (B). The asterisk(*) indicates significantly larger within 8/27 options at p ! .05. Thearrow (^) indicates significantly larger within visual images/verbal textat p ! .05.

263) p .79, p p .37), suggesting that the visual preferenceis very strong. Additionally, there is no effect of categorytype on choice of number of options (8 options p 66.2%;27 options p 33.8%; x2(1, N p 263) p .03, p p .87).Going forward, we report the results across product cat-egories except where there were significant differences.

Individual differences in processing as measured by theChilders et al. scale (1985) influence the choice of presen-tation format (B p �.06, Wald x2 p 10.34, p p .001),whereby those who are more verbal processors and lessvisual processors are less likely to have selected the visualpresentation format. There is no effect of this individualdifference variable on choice of number of options (B p�.009, Wald x2 p .28, p p .59).

Assortment Perceptions. In terms of perceptions of variety(a(3) p .81) there is a main effect of number of options inthe obvious direction (large 1 small; M27 options p 7.4, M8 options

p 6.4; F(1, 262) p 22.59, p ! .001). This is also the casefor ratings of complexity (a(3) p .87; M27 options p 5.9, M8options

p 4.3; F(1, 262) p 36.88, p ! .001). As hypothesized, selectedpresentation format also affects both of these ratings withvisual presentation leading to higher ratings of variety (Mvisual

p 7.0, Mverbal p 6.3; F(1, 262) p 28.10, p p .001) andcomplexity (Mvisual p 5.3, Mverbal p 4.0, F(1, 262) p 29.33,p ! .001), supporting hypotheses 5 and 6. An analysis ofmeans reveals that the difference in both variety and com-plexity ratings between 8 and 27 options is driven by thevisual image conditions (variety: F(1, 169) p 28.8, p !

.001; complexity: F(1, 169) p 62.7, p ! .001) and not theverbal text conditions (variety: F(1, 92) p 2.5, p p .12;complexity: F(1, 169) p .48, p ! .49). Additionally thereis an interaction effect of number of options and presentationformat on complexity (F(1, 262) p 15.07, p ! .001) butnot on variety (F(1, 262) p .98, p p .32). This offerssupport for hypothesis 6. An analysis of means reveals thatthe effect of presentation format on measures of complexityis greater among 27 options (Mvisual p 6.7, Mverbal p 4.1;F(1, 88) p 39.7, p ! .001) than among 8 options (Mvisual p4.6, Mverbal p 3.9; F(1, 173) p 6.5, p p .01) though bothare significant. (See fig. 1A and B for means by condition.For product category differences, see table 1.) Processingstyle is not a significant covariate in predicting either variety(F(1, 262) p .04, p p .90) or complexity ratings (F(1,262) p 2.67, p p .36), which is to be expected givenprocessing style influenced assortment choice.

Assortment Choice Motives. Participants who selected vi-sual presentation were more likely to say this was drivenby the motive to make the task easier, quicker, more enjoy-able, and to see the most variety (all x2(1, N p 263) 1 4.5,p ! .05). Meanwhile, those who selected verbal presentationwere more likely to say they did this in order to see themost information (x2(1, N p 263) p 17.9, p ! .001) andallow for a more precise examination (x2(1, N p 263) p22.5, p ! .001). Participants who selected the smaller choiceset were more likely to claim they were motivated by want-ing to make the task easier (x2(1, N p 263) p 44.8, p !

.001), quicker (x2(1, N p 263) p 14.9, p ! .001), andbecause they knew they would need to spend a great dealof time (x2(1, N p 263) p 18.2, p ! .001). In contrast,those who selected 27 options were, not surprisingly, morelikely to say that their choice was motivated by a desire toview more variety (x2(1, N p 263) p 124.8, p ! .001) butalso to get the most information (x2(1, N p 263) p 20.1,p ! .001) and to make the best choice (x2(1, N p 263) p33.2, p ! .001; see table 2).

Discussion

We find support for hypothesis 1; there is a general pref-erence for visual presentation over verbal when options areunfamiliar, and this preference exists across choice set sizesand product categories. Moreover, while the individual dif-ference measure of visual versus verbal processing doesinfluence presentation choice, the effect is not strong (B p





TABLE 1

STUDY 1: RATINGS OF VARIETY AND COMPLEXITY BY PRODUCT TYPE

Chose 8 items Chose 27 items

Chose verbalwords

Chose visualimages

Chose verbalwords

Chose visualimages

Crackers:N 35 56 10 36Variety 5.6* 6.6* 7.4 7.5Complexity 3.9* 5.2* 3.9* 6.9*

Mutual funds:N 28 56 20 22Variety 3.8 4.0 4.2* 6.5*Complexity 6.6 6.6 5.4* 8.0*

*Indicates significantly different at p ! .05.

�.06), again illustrating the widespread nature of the pref-erence. Ratings of variety and complexity provided initialsupport for hypotheses 5 and 6, though without randomassignment. Moreover, because respondents were able toselect their own choice set, these results are particularlynoteworthy in that prior research suggests that participantswould be confident in their choices and report high percep-tions of the choice sets to confirm their decisions (Lichten-stein, Fischhoff, and Phillips 1982; Ronis and Yates 1987).Our findings on these two measures appear to be universalwith no significant effects of individual differences in pro-cessing style. The results suggest that visual presentation ofchoice sets consistently leads to higher perceptions of bothvariety and complexity but that a greater number of optionsonly leads to greater perceptions of variety and complexitywhen the options are presented visually in images and notwhen presented verbally in text. In other words, presentationformat has a greater influence on these assortment percep-tions than does increasing the actual amount of assortmentwhile maintaining verbal presentation.

The results from the motives questions suggest that, inline with prior work on gestalt versus piecemeal processing,consumers perceive visual presentation to be an easier, faster,and more enjoyable format and that verbal presentation isbetter suited for a precise examination. If this is the case,then when visual presentation does not seem as if it willoffer the easier, faster, and more enjoyable choice experi-ence, the preference should decrease. We explore this in thenext study.

STUDY 2: ATTENUATING THEPREFERENCE FOR VISUAL

PRESENTATION

To test hypothesis 2 in study 2 we manipulate the easeof the choice process. We examine this both holding infor-mation constant and also with verbal presentation not onlyoffering greater discernment but also more information thanvisual as a stronger manipulation of this same construct. Weexpect that when choice with visual presentation is manip-

ulated to seem less easy, the preference for visual presen-tation will decrease.

Method

One hundred seventy-one participants (43% male, Mage p35) who were recruited using Amazon’s Mechanical Turkparticipated in the experiment online for compensation. Re-spondents were randomly assigned to one of six conditionsin a 3 (perceptual similarity of the nail polish colors: control/not similar, similar, versus similar-explained) by 2 (choiceset size: 8 vs. 27 options) between-subjects design.

Respondents were told to imagine they had decided topurchase nail polish as a small present for a female friendand were shopping at an online beauty supply store. Theywere told the website offers a variety of nail polishes andthat the options vary on the shape of the bottle, the color,and the type of glitter. Respondents were shown a key thatpresented the information both verbally in words and vi-sually in pictures. For the control color condition there werethree easily distinguishable colors—a blue, a red, and a pink.In both the similar color and similar-explained color con-ditions the three colors were all shades of pink, which weredifferent from each other but were more difficult to dis-criminate. In the control and similar color conditions theverbal description of the names offered basic color infor-mation (busy blue, ballet pink, and ruby red in the controland bold pink, perfect pink, and softest pink in the similarcolor conditions). We chose names in the verbal conditionthat, while different, did not offer much information aboutthe differences in the hues to hold information about therelative “pinkness” of the color constant. In contrast, in thesimilar-explained color condition, the names were 001 mostpale pink, 002 light pink, and 003 medium pink, offeringinformation on the shade of pink (see app. B, fig. B1). Afterviewing the key, participants were then told that they wouldbe shown either 8 or 27 options, depending on the condition.They were then told “when making your nail polish selec-tion, please indicate how you would prefer to be presentedwith the nail polish options” either in “written descriptionsof the nail polishes as expressed in words” or “visual images





TABLE 2

STUDY 1: PERCENT OF RESPONDENTS SELECTING MOTIVES FOR CHOICE OFPRESENTATION FORMAT AND NUMBER OF OPTIONS

Reason for selectionChose verbal

wordsChose visual

imagesChose 8options

Chose 27options

N 93 170 48 78Most information about options 49.5* 23.5* 9.2* 31.5*Allow for more precise examination 47.3* 18.2* 31.0 39.3Task easier 31.2* 46.5* 50.6* 9.0*To make the best choice 31.2 29.4 17.8* 52.8*Knowing I would spend a great deal of time 21.5 15.9 37.4* 12.4*Make task less complex 21.5 26.5 44.8* 3.4*To make the choice quickly 15.1* 34.1* 33.3* 11.2*Due to difficulty 9.7 9.4 6.9 4.5Task more enjoyable 7.5* 29.4* 15.5 16.9To view most variety of options 9.7* 20.6* 5.2* 70.8*

*Indicates significantly different within presentation format or number of options comparison at p ! .05.

of the nail polishes as shown in pictures” and were shownthree examples of each. They were then shown the choiceassortment they had selected—8 (27) options presented vi-sually (verbally) and were asked the same follow-up ques-tions as in study 1. Finally, participants were reminded thatthey were purchasing nail polish for a friend and were askedto make a selection. They were given the option to selectone option, multiple options, or “none of the above.” Theythen rated the relative importance of the three product at-tributes in a constant sum question.

Pretest on Ease. One hundred fifty-two participants (64%male, Mage p 28) who were taken from the same populationas the main study were randomly assigned to one of fiveconditions in a 2 (presentation form: image vs. text) by 3(perceptual similarity of the nail polish colors: control/notsimilar, similar, vs. similar-explained) design. Participantswere given the same purchase context, shown the condition-consistent 8-option assortment, and asked to rate it on twodifficult/easy 7-point scales: ability to discern the differencesbetween the products and ability make a choice from thischoice set. Averaging the two ease measures (a(2) p .89)among participants in the control conditions, the visual imagecondition is rated as easier than the text condition (Mvisual images

p 4.82, Mverbal text p 3.73, t(57) p 2.94, p p .005). In boththe similar and similar-explained conditions this is not thecase, with no difference on either measure (similar colors:Mvisual images p 3.57, Mverbal text p 3.20; t(61) p .98, p p .33;similar-explained colors: Mvisual images p 3.57, Mverbal text p 3.44;t(58) p .26, p p .77).

Results

Results of the constant sum question reveal a significantdifference in importance weights (F(2, 512) p 155.83, p! .001), whereby color (M p 53.3) is considered moreimportant than both glitter (M p 25.9; t(170) p 11.39, p! .001) and bottle shape (M p 20.87; t(170) p 11.97, p! .001). This did not vary across color conditions (F(2, 170)p .16, p p .89), choice of presentation format (F(1, 170)

p .02, p p .90) nor their interaction (F(2, 170) p .73, pp .48). This confirms that the manipulation of color con-ditions, indeed, varied ability to discriminate on an attributethat influences the choice decision. Also, across measuresthere were no significant differences between the similarcolors and similar-explained colors conditions. Thus, wecollapse across these two conditions when reporting (refer-ring to both as similar colors2).

Assortment Presentation Preference. The majority (83.6%,t(170) p 11.84, p ! .001) of respondents chose options pre-sented visually rather than verbally. Examining the per-centage selecting visual presentation format, there was amain effect of color (control colors p 98.2%, similar colors2

p 76.3%, x2(1, N p 171) p 13.3, p p .001), a main effectof number of options (8 items p 90.0%, 27 items p 76.5%,x2(1, N p 171) p 5.6, p p .02), and an interaction ofthese two (x2(1, N p 171 p 18.2, p ! .001). An analysisof means reveals that the effect is stronger within 27 items(control colors p 96.6%, similar colors2 p 65.3%, x2(1,N p 81) p 10.1, p p .002), than within 8 items (controlcolors p 100.0%, similar colors2 p 85.4%, x2(1, N p90) p 4.5, p p .03; see fig. 2). Individual differences inprocessing as measured by Childers et al. (1985) influencethe choice of presentation format (B p �1.4, Wald x2 p54.52, p p .001) whereby those who are more verbalprocessors and less visual processors are less likely to haveselected the visual presentation format. There was no effectof gender on choice of presentation form (visual choice:male p 86.4%, female p 80.4%, x2(1, N p 171) p 1.1,p p .29).

Assortment Perceptions. As in study 1 we examine theperceptions of both variety and complexity and how thesediffer across the resulting choice sets. Because the numberof participants selecting verbal is so low, once we crossthis by color condition and number of options, the verbaltext conditions range in size from N p 0 to N p 10.Therefore, finding any significant differences betweengroups would be noteworthy not only (as in study 1) be-





FIGURE 2

STUDY 2: CHOICE OF VISUAL REPRESENTATIONDECREASES WHEN VERBAL REPRESENTATION

DOES NOT OFFER GREATER EASE

* Indicates significantly larger within 8/27 at p ! .05.

cause the presentation form is their choice but also becausewe are considering very small sample sizes. When ex-amining perceptions of variety (a(3) p .78) there are nomain effects (all F 1 .10). When examining perceptionsof complexity (a(3) p .89), the results replicate the find-ings from study 1 whereby image presentation leads tohigher ratings of complexity (Mvisual images p 5.0, Mverbal text

p 3.1; F(1, 170) p 10.4, p p .002), particularly among27 options (marginal interaction term: F(1, 170) p 3.1, p! .10; see table 3 for full results and descriptions of maineffects).

Assortment Choice Motives

Across color conditions, participants who selected visualpresentation were more likely to say this was driven by themotive to make the task easier (x2(1, N p 171) p 5.5, pp .02), more enjoyable (x2(1, N p 171) p 16.3, p !

.001), more variety (x2(1, N p 171) p 4.0, p ! .04), lesscomplex (x2(1, N p 171) p 4.3, p p .04), and (margin-ally) quicker (x2(1, N p 171) p 3.2, p p .08). Meanwhile,those who selected verbal presentation were more likelyto say they did this in order to see the most information(x2(1, N p 171) p 45.0, p ! .001) and allow for a moreprecise examination (x2(1, N p 171) p 4.4, p p .04).These results replicate those from study 1 and support ourtheory. There were no significant differences in motivesselected across color conditions (all p 1 .10; see table 4for full results).

Choice Overload

A binary logistic regression with color conditions, numberof options, visual preference, and all of their interactions ascovariates reveals there is a marginal effect of number ofoptions (Wald x2 p 3.51, p ! .10) whereby participants whosaw 8 options were less likely to opt out of choice (3.3%)than those who saw 27 options (14.81%, x2(1) p 7.02, p p.01). Analysis of values across cells reveals that across thethree color conditions there is a trend whereby participantswho saw 27 options and selected visual presentation weremost likely to opt out of choice. While this pattern is direc-tional, it is not significant, likely due to the small sample sizesfor some cells. It is evident that this result is driven by thevisual image conditions and not the verbal text conditionsproviding initial support for hypothesis 7A (see table 3).

Discussion

Using the visual category of nail polish, it is not surprisingthat we saw an even stronger preference for visual presen-tation than in study 1. Yet, within this category, we wereable to attenuate this preference by decreasing the ease withdiscernment on the most important attribute could occur withvisual presentation. This supports hypothesis 2. While thegreater ease of processing that accompanies image depictionproduces a general sense of an easier choice process (asevidenced in the choice motives results), other factors, suchas ease of attribute level differentiation, can overcome thiscausing image depiction to lose its preferred status. Nev-ertheless, even in the close color conditions, the majority ofrespondents opted for visual presentation. We also includedtwo text conditions, one in which color discernment wasgreater in text than in visual and one where the text alsooffered additional informative cues in the form of numericalratings and useful names. Unexpectedly, there were no dif-ferences between these two conditions.

We asked respondents to evaluate their selected choiceset and, as in study 1, a larger choice set leads to greaterperceptions of complexity only when options are presentedvisually. This suggests that visual representation, while morepopular, may not be optimal with large choice. The choiceoverload results seem to corroborate this idea. Although notsignificant due to uneven cell sizes, the results within thevisual presentation format suggest that, as the number ofoptions increases, participants are actually less likely to se-lect an option and more likely to opt out of choice. Thesefindings are particularly noteworthy when considered in thecontext of prior work on choice set size and choice overload.For example, Iyengar and Lepper (2000) found that increas-ing the choice set size from 6 to 24 in the context of jam,a largely visual stimulus set, led to significantly fewer peoplemaking a purchase. We find a similar result within our visualpresentation format conditions in study 2. Additionally, con-sidering the complexity ratings, we find among those whoelected visual representation, greater ratings on complexitywhen the choice set size increases from 8 to 27 options.However, among respondents who selected the options dis-





TABLE 3

STUDY 2: RATINGS OF VARIETY AND COMPLEXITY AND PARTICIPANTSOPTING OUT OF CHOICE

8 options 27 options

Chose visualimages

Chose verbaltext

Chose visualimages

Chose verbaltext

Total across conditions:N 81 9 62 19Variety rating 6.3 6.2 6.9 6.7Complexity rating 3.9 3.4 6.2** 3.8Percent opting out of choice 3.70 0 19.4** 0

Control condition:N 28 0 28 1Variety rating 6.8 . . . 7.1 8.7Complexity rating 3.5 . . . 6.2** 2.0Percent opting out of choice 0.0% . . . 21.4% 0.0%

Similar colors:N 33 5 15 8Variety rating 5.9 6.3 6.0 6.6Complexity rating 4.0 3.6 6.1** 3.4Percent opting out of choice 9.1 0.0 13.3 0.0

Similar-explained colors:N 20 4 19 10Variety rating 6.3 6.0 7.2 6.6Complexity rating 4.4 3.1 6.2** 4.4Percent opting out of choice 3.7 0.0 19.4 0.0

NOTE.—No main effects on variety rating: color condition (Mcontrol colors p 7.5, Msimilar colors2 p 6.4; F(1, 170) p2.7, p p .11), number of options (M8 options p 6.3, M27 options p 7.3; F(1, 170) p 1.6, p p .20), visual preference(Mvisual images p 6.7, Mverbal text p 7.2; F(1, 170) p .83, p p .36). Main effect on complexity rating: no effectof color conditions (Mcontrol colors p 3.9, Msimilar colors2 p 4.4; F(1, 170) p 2.0, p p .16), main effect of numberof options (M8 options p 3.7, M27 options p 4.6; F(1, 170) p 24.8, p ! .001), main effect visual preference(Mvisual images p 5.0, Mverbal text p 3.1; F(1, 170) p 10.4, p p .002), marginally significant interaction of numberof options and visual preference (F(1, 170) p 3.1, p ! .10).

**Indicates significantly different within number of options at p ! .05.

played verbally, we find no evidence of the Iyengar andLepper overload and the pattern of results suggests that witha larger sample size we might see no difference or even thereverse whereby perceived complexity decreases with anincrease from 8 to 27 options.

When asked to indicate their motives for selecting visualor verbal presentation format, again the results replicatethose from study 1. Consumers perceive visual presentationto be an easier, faster, and more enjoyable format and verbalpresentation to be better suited for a precise examination.We hypothesize that the ease, speed, and enjoyment thatcomes with gestalt processing may not be ideal with largerchoice sets and that this type of processing that occurs withimages may be causing the greater perceptions of complexitywhen considering 27 options. In the next study we use eye-tracking devices to specifically examine the processing thatnaturally occurs with visual versus verbal stimuli.

STUDY 3: EXAMINING DIFFERENCES INEYE MOVEMENTS ACROSS CHOICE SETS

In study 3 we use eye-tracking technology to examinethe effect of processing directly and test hypothesis 3. Byusing eye-tracking devices, we are able to measure theamount of time spent evaluating each option, whether each

option is examined, and the pattern of the fixations. Withrespect to these three variables, we predict that regardlessof choice set size, participants will spend longer evaluatingeach option when presented with verbal depictions thanwhen presented with visual depictions. With respect to thevisual movement path and number of options skipped, insmall choice sets (where there are fewer degrees of freedomin terms of visual movement paths), we predict there willbe no differences between visual and verbal depictions ineither the number of options examined or in movement pat-terns. However, in larger choice sets, we expect respondentswill view verbal depictions in more systematic/reading-likemovement patterns and will examine a greater percentageof the options within the choice set.

In this (and the next two studies) we use crackers as theproduct category. The set-up of study 3 is similar to thecracker condition of study 1. However, one fundamentaldifference is that, whereas in the prior studies participantschose their choice set, to ensure against self-selection con-founds, going forward we randomly assign respondents toconditions.

MethodForty students (Mage p 21; 70% female) at University of

Technology, Sydney, participated in the experiment online





TABLE 4

STUDY 2: PERCENT OF RESPONDENTS SELECTING MOTIVES FOR CHOICE OFPRESENTATION FORMAT ACROSS COLOR CONDITIONS

Total Control Similar colors Similar-explained colors

Reason for selection

Choseverbalwords

Chosevisual

images

Choseverbalwords

Chosevisual

images

Choseverbalwords

Chosevisual

images

Choseverbalwords

Chosevisual

images

N 28 143 1 56 13 48 14 39Most information about options 71.4* 14.7* 0.0 17.9 76.9* 10.4* 71.4* 15.4*Allow for more precise

examination 60.7* 41.3* 100.0 48.2 53.8 33.3 64.3 41.0To make the best choice 42.9 32.2 0.0 30.4 38.5 27.1 50.0 41.0Task easier 25.0* 47.6* 0.0 60.7 23.1 39.6 28.6 38.5To make the choice quickly 31.4** 37.8** 0.0 35.7 30.8 43.8 14.3 33.3Knowing I would spend a great

deal of time 21.4 14.0 0.0 14.3 23.1 16.7 21.4 10.3To view most variety of options 10.7* 28.7* 0.0 21.4 15.4** 41.7** 7.1 23.1Make task less complex 10.7* 28.7* 0.0 30.4 15.4 27.1 7.1 28.2Task more enjoyable 3.6* 42.7* 0.0 35.7 0.0* 56.3* 7.1* 35.9*Due to difficulty 0.0 1.4 0.0 0.0 0.0 4.2 0.0 0.0

*Indicates significantly different within color condition at p ! .05.**Indicates significantly different within color condition at p ! .10.

for payment. The design was a 2 (number of choice options:4 vs. 14) by 2 (presentation form: image vs. text) between-subjects design.

Throughout the studies participants were seated one at atime in front of a computer equipped with Tobii Studio 2.2technology. Participants were presented with the crackerschoice set commensurate with their condition twice: firsttime to look over the choice set and then to select one option,more than one option, or “none of the above.”

Commensurate with the necessities of an eye-trackingstudy, there were several differences between this study andprior and subsequent studies. First, a lab assistant neededto calibrate the participants one by one, so the eye move-ments could be tracked by the software. Second, the labassistant continued to monitor the participants one on one,as they proceeded through the experiment. Finally, while inthe other studies the presentation of the particular (cracker)options within each choice set was randomized betweensubjects, in this study they were not. This allowed for re-cording of which specific option was being examined at anypoint in time. The choice set for all four versions weredelineated into specific areas of interest (AOIs), and eachAOI was associated with a single cracker stimulus. Theresults are analyzed such that all points within the AOI areconsidered a fixation on that specific cracker option. Usingthe Tobii fixation filter (Tobii 2010), we compare patternand durations of fixations across the experimental condi-tions.

In line with previous research we distinguish betweensystematic examination (defined as similar to reading) andnonsystematic examination (Hogeboom and van Leeuwen1997). To do this we calculated the number of movementsbetween crackers, whereby the participant’s eye fixationsmoved from one cracker AOI to a neighboring cracker AOI(either right, left, up, or down but not diagonally) and also

included movement from the end of one row to the begin-ning of another. From this we calculated a systematic-move-ment percentage, taking into account the total number ofcrackers examined; if participants did not look at a crackerat all, it was not included in this calculation.

Results

Initial Presentation. We examine the manner in whichparticipants processed the options when initially presentedwith the choice set. Looking at amount of time taken toview each option, in both the large (F(1, 278) p 7.65, p !

.006) and small (F(1, 78) p 13.5, p ! .001) choice sets,respondents take longer with verbal than visual stimuli. Onour other measures overall we see a general pattern, wherebythere are no differences within the small choice sets, but inthe larger choice sets examination of the options is lesssystematic with visual than with verbal presentation (seetable 5).

Choice. Respondents were then presented again with thechoice set in order to make a choice. All respondents exceptone (39 of 40) made at least one choice from the selection(F(3, 36) p 1.00, p p .40); thus we do not replicate study2 results nor find support for hypothesis 7A. We attributethis to the experimental setting necessary for eye tracking;each participant is alone in the lab with the lab assistant,and probably demand effects motivate the participant tomake a choice and not select “none of the above.” Withrespect to the manner in which participants processed theoptions, as with the initial view we see no differences withinthe smaller choice sets in terms of systematicity (Mverbal p100%, Mvisual p 100%; p 1 .10) or number of optionsskipped (Mverbal p .20, Mvisual p .80; F(1, 19) p 2.46, p p.14). In contrast, in the larger choice sets participants who





TABLE 5

STUDY 3: EYE-TRACKING PROCESS MEASURES IN INITIAL CHOICE SETVIEWING, CHOICE, AND MATCHING


Verbal text Visual images Verbal text Visual images

Initial view:Time taken on each option viewed 5.21* 2.47 1.91* 1.46Percent systematic movements 100.0 100.0 77.9 68.6Number of options skipped 0 0 0.5 0.3

Choice:Time taken on each option viewed 8.63** 4.31 18.69 13.86Percent systematic movements 100.0 100.0 69.5 34.8Number of options skipped 0.2 0.8 1.3 9.7*

NOTE.—Time statistics given in seconds.*Indicates significantly larger within 8/27 options at p ! .05.**Indicates significantly larger within 8/27 options at p ! .10.

saw verbal descriptions are more methodical and careful intheir examination of the options (percent systematic: 69.5%;Moptions skipped p 1.3) than participants who saw visual de-scriptions (percent systematic: 34.8%; F(1, 19) p 18.1, p! .001; Moptions skipped p 9.7; F(1, 19) p 20.3, p ! .001). Wealso find that respondents take longer with text than imagesthough marginal in the smaller choice sets (Mverbal p 8.5,Mvisual p 4.3; F(1, 19) p 3.63, p p .07) and only directionalin the larger choice sets (Mverbal p 18.7, Mvisual p 13.9; F(1, 19) p 1.15, p p .30; see table 5). These results are allconsistent with hypothesis 3.

Discussion

Participants are generally systematic in their examinationof options in the small choice sets regardless of depictionstyle, but for large choice sets, they are more systematic andthorough when options are verbally displayed compared tothe visual display. We observe that for large assortments thepiecemeal processing of individual verbal options carriesover to produce a slower and more systematic processingof the overall choice set while the natural gestalt processingof individual visual stimuli leads to a faster and more hap-hazard processing of the assortment. In the next two studieswe examine the perceptual and behavioral consequences ofthese differences in processing. In study 4 we considerwhether the more systematic processing that occurs in largechoice sets with text as opposed to images is more effective.

STUDY 4: VISUAL VERSUS VERBALPRESENTATION IN A MATCHING TASK

Study 4 is designed to test the effectiveness of processingthat naturally occurs during choice from either a visually orverbally depicted assortment. One way to measure process-ing is through performance on a matching task (Adams1967; Roediger and Guynn 1996). Therefore, we use a sur-prise matching task after a choice task to measure recog-nition and retention of items as an indication of the pro-cessing that has occurred. Per hypothesis 4 we expect that

with a large choice set performance on the matching task(number of options correctly recalled and recognized) willbe better when the options were presented verbally thanwhen they were presented visually. We expect no such effectwith a smaller choice set. However, previous research iden-tifying the picture superiority effect predicts that with singleitems, performance on a matching task will be better withvisual presentation than verbal (Childers and Houston 1984;Nelson, Reed, and Walling 1976). We therefore expect visualdepiction to result in better performance in a smaller choiceset. Therefore, we are predicting a crossover interaction.After the matching task we also ask participants how surethey were of their responses and whether they would haveperformed better or worse if they had done the task withthe alternate presentation format. Our theory suggests a maineffect here regardless of choice set size: participants whosaw images versus words will feel more certain since thetask will feel easier. Similarly, all participants will preferthe task with images, regardless of which they saw.

Method

Two hundred and two students (Mage p 22.3; 60% female)at the Wharton School, University of Pennsylvania, partic-ipated in the experiment online for payment. The design wasa 2 (number of options: 4 from 8 vs. 14 from 27) by 2(presentation form: image vs. text) between-subjects design.In this and the next study, the full 8- and 27-item assortmentswere full factorial designs for three attributes (shape, flavor,or topping) and varied in the number of attribute levels (twoor three).

Participants were first shown an information key (see app.A, fig. A1) and asked to imagine they were shopping forcrackers for a party. They were then shown the assortmentof crackers commensurate with the condition (either 4 or14, presented either in images or text). After making achoice, respondents were shown the complete set, all 8 or27, and were asked to select the 4/14 that were shown pre-viously. Respondents were then asked how sure they wereof their response using a 100-point sliding scale and to rate





FIGURE 3

STUDY 4: NUMBER OF OPTIONS CORRECTLY IDENTIFIEDIN SURPRISE MATCHING TASK

** Indicates significantly larger than visual images/verbal wordswithin 8/27 options at p ! .05.

on a 7-point scale if they had seen written descriptions/visualimages (instead of visual/written) would they have per-formed worse or better. Participants then gave their genderand age and were thanked.

Results

Performance on Matching Task. Considering number ofcorrect options selected, we find no main effect of presen-tation format (Mverbal p 5.1, Mvisual p 5.2; F(1, 201) p .09,p p .76), a main effect of number of options (M8 options p2.4, M27 options p 7.8; F(1, 201) p 1,251.91, p ! .001), anda significant interaction of these two variables (F(1, 201) p57.82, p ! .001). Among 8 options, respondents who sawimages perform better than those who saw verbal descrip-tions (Mverbal p 1.9, Mvisual p 2.9; F(1, 97) p 46.59, p !

.001). However, among 27 options, respondents who sawimages performed worse than those who saw verbal de-scriptions (Mverbal p 8.1, Mvisual p 7.5; F(1, 102) p 6.14,p p .02; see fig. 3). This supports hypothesis 4.

Perceptions of Matching Task. On both measures the re-sults are as expected; participants who saw images weremore sure of their response than those who saw text (Mverbal

p 33.0, Mvisual p 47.7; F(1, 201) p 13.42, p ! .001), andthis is driven by participants who saw 8 options (F(1, 97)p 21.54, p ! .001) rather than 27 (F(1, 102) p .75, p p.39). Examining responses to whether their performancewould be better with the alternate presentation format re-spondents who saw images are less likely to say that theywould do better if they had seen words than vice versa(Mverbal p 5.7, Mvisual p 4.3; F(1, 201) p 45.74, p ! .001;see table 6 for full results and all main effects).

Discussion

In line with prior work on the picture superiority effect(Nelson et al. 1976), when the choice set is small, imagesproduce greater performance than text. Counter to past re-search, however, we find that with a larger choice set theopposite is true. The results of study 4 corroborate the im-plications from study 3 that for smaller choice sets imageslead to better processing but that for larger choice sets verbaldescriptions are better. Despite this, regardless of choice setsize, respondents presented with verbal descriptions weremore likely than those who saw images to say the matchingtask would have been easier had they been presented withinformation in the alternate format. This speaks to thestrength and overuse of the visual preference heuristic—apparent even after a task where verbal presentation led tosuperior performance.

The matching task could be considered a model of howconsumers might search if they had strong preexisting pref-erences and know what they are seeking. In such a case likesmall choice sets, visual presentation could be consideredoptimal. As such, this study offers a potential moderatorbeyond the categorical ones suggested in studies 1 and 2,when visual presentation is optimal.

In the next and final study we consider further conse-quences of the differences in processing observed in study3; study 5 examines the perceptual differences (variety andcomplexity), as well as behavioral differences (choice over-load) hinted at in the first two studies.

STUDY 5: VISUAL VERSUS VERBALPRESENTATION FORMAT, CHOICE SET

PERCEPTIONS, AND CHOICE OVERLOAD

In study 5 we test hypothesis 7A on choice overload. Wealso examine hypotheses 5 and 6 on perceptions of varietyand complexity and consider how the latter may influencechoice behavior per hypothesis 7B. We predict that the fasterand more haphazard examination of the assortment withvisual depiction that we observed in study 3 and seems lesseffective per study 4 is insufficient in larger choice sets;consumers thus perceive the choice set to be highly complexand will be more likely to opt out of choice all together.However, the more systematic approach with verbal versusvisual presentation allows participants to more carefully pro-cess the options, which can mitigate perceptions of com-plexity; thus in the verbal condition, the choice deferralshould not be as great.

Additionally, we include two new scales, the 13-itemmaximizer/satisficer scale (Schwartz et al. 2002) that iden-tifies whether someone tends to seek out the first option thatcrosses the threshold of being acceptable (satisficer) or mustfind the best available option and thus will be more nega-tively influenced by large choice sets (maximizer) and the11-item centrality of visual product aesthetics (CVPA) scale





TABLE 6

STUDY 4: RESPONSE TO MEMORY TASK


Verbaltext

Visualimages

Verbaltext

Visualimages

Number of options identified correctly 1.9 2.9** 8.1** 7.5Sure of response to memory task 41.6 66.7** 25.3 29.0Other presentation format would be better 5.7** 4.3 5.7** 4.2

NOTE.—Sure of response to memory task: main effect of presentation format (Mverbal p 33.0, Mvisual p 47.7; F(1, 201)p 13.42, p ! .001), main effect of number of options (M8 options p 54.6, M27 options p 27.1; F(1, 200) p 56.77, p ! .001),interaction effect (F(1, 201) p 11.25, p p .001). Other presentation format better: main effect of presentation format(Mverbal p 5.7, Mvisual p 4.3; F(1, 201) p 45.74, p ! .001), no main effect of number of options (F(1, 201) p .00, p p.99).

**Indicates significantly larger within 8/27 options at p ! .05.

(Bloch, Brunel, and Arnold 2003) that measures how im-portant the look of a product is to a consumer.

Method

Two hundred thirty-six undergraduate business students(Mage p 18.6; 61% female) at the University of Miami par-ticipated in the experiment online for course credit. Thedesign was a 2 (number of choice options: 8 vs. 27) by 2(presentation form: image vs. text) between-subjects design.

Participants were first shown an information key (see app.A, fig. A1) and asked to imagine they were shopping forcrackers for a party. They were then shown the assortmentof crackers commensurate with the condition. Respondentsrated the assortment on the variety and complexity as instudies 1 and 2. They were then shown the assortment againand asked to make a selection where they were allowed tochoose one option, more than one option, or “none of theabove.” Participants then filled out the individual differencescales, gender and age.

Results

Assortment Perceptions. We find a main effect of numberof options on perceptions of variety (a(3) p .83; F(1, 234)p 27.5, p ! .001) and complexity (a(3) p .83; F(1, 234)p 37.7, p ! .001). As expected, respondents perceivedgreater variety and complexity for the 27-option than forthe 8-option assortment. There is also a main effect of pre-sentation form with higher scores for images than text onmeasures of variety (F(1, 234) p 10.8, p p .001) andcomplexity (F(1, 234) p 3.99, p p .047). There is also asignificant interaction for both variety (F(1, 234) p 27.48,p ! .001) and complexity (F(1, 234) p 36.42, p ! .001).

Supporting hypothesis 5, participants who saw visual im-ages perceive greater variety than those who saw verbal textdescriptions for both 8 options (Mvisual p 6.5, Mverbal p 5.7;F(1, 115) p 6.88, p p .01) as well as 27 options (Mvisual

p 7.4, Mverbal p 5.9; F(1, 117) p 21.22, p ! .001), althoughthe difference is greater for 27. There is no difference inperceptions of variety within verbal text for 8 and 27 options(F(1, 127) p .41, p p .52), but there is for visual images

(F(1, 105) p 17.03, p ! .001). The differences in perceivedcomplexity between presentation formats are driven by dif-ferences when respondents are presented with 27 options(Mvisual p 6.1, Mverbal p 4.6; F(1, 117) p 16.04, p ! .001).This supports hypothesis 6. There was no significant dif-ference when respondents were presented with 8 options(Mvisual p 4.2, Mverbal p 4.0; F(1, 115) p .375, p p .54;see fig. 4A and B).

Choice Overload. We find no main effect on likelihoodto opt out of choice for the number of options (8 options p9%, 27 options p 8%, x2(1) p .002, p p .97) nor forpresentation format (visual p 10%, verbal p 7%, x2(1) p.82, p p .36). There is, however, a significant interactioneffect (x2(1) p 4.36, p p .04). There is no difference inthe number opting to select none of the above among 8options (visual p 5%, verbal p 11%, x2(1) p 1.30, p p.24), while there is a difference in the opposite directionamong 27 options (visual p 16%, verbal p 3%, x2(1) p6.15, p p .01) with more respondents opting out of choicewho saw images than saw text (fig. 5). This supports hy-pothesis 7A. Looking only at the results for respondentswho saw the options displayed visually, we replicate theresults of Iyengar and Lepper (2000). We find that there isgreater overload when the choice set size increases from 8to 27 options consistent with greater ratings on complexity(F(1, 105) p 34.71, p ! .001), as well as a marginal increasein choice of “none of the above” (x2(1) p 3.09, p p .08).However, when we examine the results for respondents whosaw the options displayed verbally, this is not the case; thereis no significant difference in ratings of complexity (F(1,127) p 2.81, p p .10). Moreover, the percent selecting“none of the above” actually decreases marginally (x2(1) p3.61, p p .06). Neither of the individual difference measures(maximizing vs. satisficing and CVPA) revealed significantmain effects nor interaction effects when we included themin the regression (all Wald’s ! 1.3, all p 1 .10).

Mediation Analyses. We used a bootstrapping approachto test for mediation (Preacher and Hayes 2004; Zhao,Lynch, and Chen 2010). First we examined the interactioneffect of presentation format and number of items on choice





FIGURE 4

STUDY 5: PERCEIVED VARIETY AND COMPLEXITY

NOTE.—Perceived variety (A) and perceived complexity (B).Asterisks (**) indicate significant differences between 8 and 27 options(main effect), visual and verbal (main effect) as well as within 8 optionsbetween visual and verbal and within 27 options between visual andverbal at p ! .05. The plus sign (�) indicates a significant differencewithin visual for 8 and 27 options at p ! .05.

of “none of the above” and tested for complexity (averageof three measures) as a mediator. We found that the meanindirect effect from the bootstrap analysis is positive andsignificant (2.7), with a 95% confidence interval excludingzero (.72 to 10.3). The direct effect (1.4) is not significant(p p .16), indicating indirect-only mediation. This supportshypothesis 7B that it is high perceptions of complexity thatdrive consumers to opt out of making a choice. We alsotested variety (average of three measures) as a potentialmediator, and in line with our characterization of variety itdoes not mediate the effect; the 95% confidence interval forthe indirect effect includes zero (�.74 to 1.33).

Discussion

We find support for hypotheses 5, 6, 7A, and 7B. Despiterespondents showing a strong preference for visual ratherthan verbal descriptions (studies 1 and 2), in larger choicesets visual presentation format causes greater perceptions ofcomplexity and thus greater choice overload. Study 5 there-fore identifies a moderator for the choice overload effect

that Iyengar and Lepper identified. It is due to the visualnature of the stimuli, which facilitated a more gestalt-likeprocessing of the options, whereas the more piecemeal-likeprocessing style that results from verbal depiction allowsfor less choice deferral.

GENERAL DISCUSSIONIn this research, we find evidence of a visual preference

heuristic such that consumers prefer visual rather than verbaldepiction of stimuli in an assortment, regardless of choiceset size. This occurs even with product categories not typ-ically depicted visually and for which there is no naturalvisual representation (mutual funds in study 1) and evenwhen verbal depiction allows for greater differentiation(close color conditions in study 2). However, we find thatthis preference for visual can lead to suboptimality. Whenchoice sets are small, visual presentation results in moreefficient processing of stimuli, increased perceived variety,and faster processing times. However, when choice sets arelarge, visual presentation motivates a facile approach andwhen preferences are unknown consumers may do betterwith verbal depiction that induces a slower, more systematicprocessing style. We provide evidence for our conclusionsthrough five studies.

In study 1 we showed that consumers defer to a generalvisual preference heuristic; they opt for visual rather thanverbal depiction regardless of choice set size. We did notfind a significant difference in this preference as a functionof product class, although we predicted we would. In study2, we varied the ability to differentiate among attribute levelsusing diverse or similar colors of nail polish. Again thepreference for visual depiction still existed, but we find asignificant decrease in this preference when attribute dif-ferentiation was difficult, suggesting perceptions of ease asa driver of the heuristic. In study 3 we used eye-trackingtechnology to examine the actual process with which optionsare examined. Across conditions participants spent longerexamining each option with verbal presentation than visual,indicative of piecemeal versus gestalt processing. We alsoobserved how this difference in processing the individualitems influenced processing of the entire choice set. In smallchoice sets where gestalt processing is adequate we foundno differences in assortment processing between presenta-tion formats. However in larger choices sets the piecemealprocessing of verbal options caused respondents to moresystematically process the entire choice set, scanning in apattern similar to reading and skipping fewer options thanwhen presented with visual depiction. Study 3 therefore pro-vided evidence of the two differences in processing, bothat the individual item level and at the assortment level. Wesuggest that these natural differences in processing of imagesversus words results in both the “visual preference heuris-tic,” as well as the negative impact of visual depiction onthe choice process in larger choice sets.

The next two studies looked more specifically at this neg-ative impact. In study 4 we used a surprise matching taskas a measure of processing effectiveness. We found that for





FIGURE 5

STUDY 5: PARTICIPANTS OPTING OUT OF CHOICE—SELECTING “NONE OF THE ABOVE”

NOTE.—Asterisks (**) indicate significant difference betweenvisual and verbal within 27 options (main effect) at p ! .05. The plussign (�) indicates significant difference between 8 and 27 optionswithin visual at p ! .10. The arrow (^) indicates significant differencebetween 8 and 27 options within verbal at p ! .10.

small choice sets, visual depiction was better for a matchingtask than verbal depiction. However, these results reversefor large choice sets where verbal depiction yielded betterresults. Study 5 revealed that visual presentation led to pos-itive perceptions of more variety for small choice sets anda greater likelihood to choose. In large choice sets, however,this increased perception of variety was accompanied bynegative perceptions of complexity resulting in a decreasedlikelihood to choose. Thus, studies 4 and 5 suggest that thevisual preference heuristic is overapplied; in studies 1 and2 participants opted for visual presentation even with largechoice sets while studies 4 and 5 revealed this not to beoptimal. However, the choice sets we used in all our studiesrepresented situations where the options were all new andthus preferences unknown. In situations where the choicesets are familiar or where satisficing is acceptable the effi-cient, fast gazing strategy applied to visual assortmentsmight not only suffice but be preferable.

We limited our exploration to the use of words or imagesexclusively and did not consider when a combination of thetwo might be used. Future research might tackle these ques-tions. Additionally, while prior work on choice overload hasconsidered the organization of the stimuli (e.g., Kahn andWansink 2004), and the trade-off of number of attributesversus attributes levels (e.g., Lee and Lee 2004; Malhotra1982) we did not. The options were always presented in arandomized fashion. Had they been presented in a moreorderly manner it is possible that the less systematic orderof processing the visual stimuli in the large choice set (study3) may have dissipated. However, it seems unlikely that thetime spent processing the options would have also increased

and, thus, our results likely still hold even with a moreorganized option presentation.

This research furthers our understanding in three distinctareas: choice overload, verbal versus visual depiction andprocessing, and the influence of processing of individualitems on that of full assortments. Specific to prior work onchoice overload, these studies provide more insight to helpexplain the findings of Iyengar and Lepper (2000). We rep-licated their results of choice deferral for visual stimuli whenour choice set was increased from 8 to 27 options. However,this was only the case when options were presented visually;when options were presented in text, this did not occur. Oneimplication of this is that if Iyengar and Lepper’s jam ex-periment was run with textual descriptions of the jams, theresults might be different. In fact, Iyengar and Lepper in-cluded a study with text-based stimuli, essay topics, andreplicated their overload results. Our results suggest that theissue of how much time was spent examining the optionsmay have had an effect; we see that respondents spend moretime with text than images. Perhaps if respondents in theessay task were forced to spend longer they would be morelikely to choose. In any event, our results would not predictthe choice overload findings that Iyengar and Lepper showfor the essay test. However, our results on the benefit ofverbal depiction for large choice sets is consistent with theresults found by Ketcham et al. (2012) where they foundthat senior citizens under Medicare Part D Plans made betterchoices (i.e., were less likely to over spend on prescriptiondrugs) when choosing from a larger rather than smaller as-sortment.

We based our hypotheses on prior work on the differencesin the natural gestalt processing of images versus the piece-meal processing necessary with words. What our research,and study 3 in particular, reveals is that these differences inprocessing of the individual items carry over and affect theway entire sets are processed. It is the ease and speed withwhich we naturally process individual images that causesus to be less ordered and deliberative when faced with alarge group of images as compared to a large group of textualstimuli. This parallels our measures of variety and com-plexity. While variety is generally positive, relates to pref-erence matching, and benefits assortment choice, complexityis the negative associate of variety and relates to evaluationof the specific options within the set. As such high varietythat attracts consumers to an assortment, particularly withvisual depiction, can be accompanied by complexity thatmakes the processing overwhelming and leads to choiceoverload.

Our results suggest that consumers do not always knowwhat is best. While visual depiction can be advantageous,there are some choice situations where consumers shouldtake more time and process more deeply. The current ten-dency for mobile apps to favor graphics exclusively mayparadoxically cause consumers to choose not to purchase,especially for unfamiliar stimuli. While the images are fun,there may be a tendency to gloss over them rather than makea purchase. In situations where marketing materials for the





outlet are separate from the actual shopping experience, forexample, the homepage for an online retailer versus theproduct selection page, it would be beneficial to have theupfront information about the store and its products pre-sented in a visual manner, emphasizing the ease of the de-cision process and amount of variety. Then, at the point ofactual consideration for purchase, on the product offeringspages a more text-based interface should cause consumersto slow down, review each option more carefully, and beless likely to opt out of choice. Alternatively, if a companywishes to reap the benefits of images without suffering fromthe less systematic processing and all of its perceptual andbehavioral consequences, then keeping the choice set seem-ingly small is optimal. Apple has always been a visuallybased company and their website does a good job of min-imizing the number of images presented at a time despitetheir large product offering. In this manner they are able tomake the choice process seem quick, fun, and easy, withoutseeming complex and overwhelming.

DATA COLLECTION INFORMATIONStudy 1: The first author collected the data online using

Amazon’s Mechanical Turk to recruit the sample. The first

author analyzed the data, and both authors reviewed anddiscussed the results. All of this took place in summer 2012.

Study 2: The first author collected the data online usingAmazon’s Mechanical Turk to recruit the sample. The firstauthor analyzed the data, and both authors reviewed anddiscussed the results. All of this took place in winter 2013.

Study 3: The second author oversaw collection of the databy the lab manager in the Behavioural Lab at University ofTechnology, Sydney. The lab manager also performed pre-liminary creation of the variables from the raw data. Thefirst author analyzed the data, and both authors reviewedand discussed the results. All of this took place in fall 2011.

Study 4: The second author oversaw collection of the databy research assistants at the Wharton Behavioral Lab, Uni-versity of Pennsylvania. The first author analyzed the data,and both authors reviewed and discussed the results Thisall took place in spring 2011.

Study 5: The first author oversaw collection of the databy research assistants at the Canes Behavioral Laboratory,University of Miami School of Business. The first authoranalyzed the data, and both authors reviewed and discussedthe results. All of this took place in spring 2011.



APPENDIX AFIGURE A1



FIGURE A2

NOTE.—Image depiction of fees varies by color.



APPENDIX BFIGURE B1





REFERENCES

Adams, John (1967), Human Memory, New York: McGraw Hill.Alba, Joseph W., and J. Wesley Hutchinson (1987), “Dimensions

of Consumer Expertise,” Journal of Consumer Research, 13(4), 411–54.

Arnheim, Rudolf (1974), Art and Visual Perception: A Psychologyof the Creative Eye, Berkeley: University of California Press.

Bawa, Kapil, Jane T. Landwehr, and Aradhna Krishna (1989),“Consumer Response to Retailers’ Marketing Environments:An Analysis of Coffee Purchase Data,” Journal of Retailing,65 (Winter), 471–95.

Birkhoff, George D. (1933), Aesthetic Measure, Cambridge, MA:Harvard University Press.

Bloch, Peter H., Frederic F. Brunel, and Todd J. Arnold (2003),“The Centrality of Visual Product Aesthetics: Concept andMeasurement,” Journal of Consumer Research, 29 (4),551–65.

Broniarczyk, Susan M., Wayne D. Hoyer, and Leigh McAlister(1998), “Consumers’ Perceptions of the Assortment Offeredin a Grocery Category: The Impact of Item Reduction,” Jour-nal of Marketing Research, 35 (2), 166–76.

Carr, Thomas H., Charles McCauley, Richard D. Sperber, and C.M. Parmelee (1982), “Words, Pictures, and Priming: On Se-mantic Activation, Conscious Identification, and the Auto-maticity of Information Processing,” Journal of ExperimentalPsychology: Human Perception and Performance, 8 (6),757–77.

Chase, William G., and Herbert A. Simon (1973), “Perception inChess,” Cognitive Psychology, 4 (January), 55–81.

Chernev, Alexander, Ulf Bockenholt, and Joseph Goodman (2010),“Commentary on Scheibehenne, Greifeneder, and Todd;Choice Overload: Is There Anything to It?” Journal of Con-sumer Research, 37 (3), 426–28.

Childers, Terry L., and Michael J. Houston (1984), “Conditionsfor a Picture-Superiority Effect on Consumer Memory,” Jour-nal of Consumer Research, 11 (2), 643–54.

Childers, Terry L., Michael J. Houston, and Susan E. Heckler(1985), “Measurement of Individual Differences in Visual ver-sus Verbal Information Processing,” Journal of Consumer Be-havior, 12 (September), 125–34.

Dhar, Ravi (1997), “Consumer Preference for a No-Choice Op-tion,” Journal of Consumer Research, 24 (September),215–31.

Fitzsimons, Gavan J., Eric A. Greenleaf, and Donald R. Lehmann(1997), “Decision and Consumption Satisfaction: Implica-tions for Channel Relations,” Working Paper No. 313, Mar-keting Studies Center, University of California, Los Angeles.

Hamilton, Rebecca, and Debora V. Thompson (2007), “Is There aSubstitute for Direct Experience? Comparing Consumers’Preferences after Direct and Indirect Product Experiences,”Journal of Consumer Research, 34 (December), 546–55.

Hart, Roderick P. (1997), “Analyzing Media,” Modern RhetoricalCriticism, 2nd ed. Boston: Allyn & Bacon, 177–208.

Hoch, Stephen J., Eric L. Bradlow, and Brian Wansink (1999),“The Variety of Assortment,” Marketing Science, 18 (4),527–46.

Hoch, Stephen J., and John Deighton (1989), “Managing WhatConsumers Learn from Experience,” Journal of Marketing,53 (April), 1–20.

Hogeboom, Menno, and Cees van Leeuwen (1997), “Visual SearchStrategy and Perceptual Organization with Individual Pref-

erence and Structural Complexity,” Acta Psychologica, 95 (2),141–64.

Holbrook, Morris B., and William L. Moore (1981), “Feature In-teractions in Consumer Judgments of Verbal versus PictorialPresentations,” Journal of Consumer Research, 8 (1), 103–13.

Hsee, Christopher K., and Yuval Rottenstreich (2004), “Music,Pandas, and Muggers: On the Affective Psychology of Value,”Journal of Experimental Psychology: General, 133 (1), 23–30.

Huffman, Cynthia, and Barbara E. Kahn (1998), “Variety for Sale:Mass Customization or Mass Confusion?” Journal of Retail-ing, 74 (4), 491–513.

Iyengar, Sheena S., and Mark R. Lepper (2000), “When Choice IsDemotivating: Can One Desire Too Much of a Good Thing?”Journal of Personality and Social Psychology, 79 (6), 995–1006.

Jacoby, Jacob, Donald E. Speller, and Carol K. Berning (1974),“Brand Choice Behavior as a Function of Information Load:Replication and Extension,” Journal of Consumer Research,1 (June), 33–42.

Kahn, Barbara E., and Donald R. Lehmann (1991), “ModelingChoice among Assortments,” Journal of Retailing, 67 (3),274–99.

Kahn, Barbara E., and Brian Wansink (2004), “The Influence ofAssortment Structure on Perceived Variety and ConsumptionQuantities,” Journal of Consumer Research, 30 (4), 519–33.

Ketcham, Jonathan, Claudio Lucarelli, Eugenio Miravete, and M.Christopher Roebuck (2012), “Sinking, Swimming, or Learn-ing to Swim in Medicare Part D?” American Economic Re-view, 102 (6), 2639–73.

Khateb, Asaid, Alan J. Pegna, Christoph M. Michel, Theodor Lan-dis, and Jean-Marie Annoni (2002), “Dynamics of Brain Ac-tivation during an Explicit Word and Image Recognition Task:An Electrophysiological Study,” Brain Topography, 14(Spring), 197–213.

Lee, Byung-Kwan, and Wei-Na Lee (2004), “The Effect of Infor-mation Overload on Consumer Choice Quality in an On-LineEnvironment,” Psychology and Marketing, 21 (3), 159–83.

Lee, Leonard, On Amir, and Dan Ariely (2009), “In Search ofHomo Economicus: Cognitive Noise and the Role of Emotionin Preference Consistency,” Journal of Consumer Research,36 (2), 173–87.

Lichtenstein, Sarah, Baruch Fischhoff, and Lawrence D. Phillips(1982), “Calibration of Probabilities: The State of the Art to1980,” in Judgment under Uncertainty: Heuristics and Biases,ed. Daniel Kahneman, Paul Slavic, and Amos Tversky, Cam-bridge: Cambridge University Press.

Lieberman, Matthew D., Ruth Gaunt, Daniel T. Gilbert, and YaacovTrope (2002), “‘Reflexion and Reflection’: A Social CognitiveNeuroscience Approach to Attributional Inference,” in Ad-vances in Experimental Social Psychology, Vol. 34, ed. MarkP. Zanna, San Diego, CA: Academic Press, 199–249.

Luna, David, and Laura A. Perrachio (2003), “Visual and Lin-guistic Processing of Ads by Bilingual Consumers,” in Per-suasive Imagery: A Consumer Response Perspective, ed.Linda M. Scott and Rajeev Batra, Mahwah, NJ: Erlbaum,153–75.

Malhotra, Naresh K. (1982), “Information Load and ConsumerDecision Making,” Journal of Consumer Research, 8 (4),419–30.

Mantrala, Murali K., Michael Levy, Barbara E. Kahn, Edward J.Fox, Peter Gaidarev, Bill Bankworth, and Denish Shah,





(2009), “Why Is Assortment Planning so Difficult for Re-tailers? A Framework and Research Agenda,” Journal of Re-tailing, 85 (1), 71–83.

Meyer, Robert J. (1986), “A Theory of the Inductive Learning ofMultiattribute Preferences,” in Perspectives on Methodologyin Consumer Behavior, ed. David Brinberg and Richard J.Lutz, New York: Springer, 253–85.

——— (1987), “The Learning of Multiattribute Judgment Poli-cies,” Journal of Consumer Research, 14 (2), 155–73.

Mogilner, Cassie, Tamar Rudnick, and Sheena Iyengar (2008),“The Mere Categorization Effect: How the Presence of Cat-egories Increases Choosers’ Perceptions of Assortment Va-riety and Outcome Satisfaction,” Journal of Consumer Re-search, 35 (2), 202–15.

Morales, Andrea, Barbara E. Kahn, Leigh McAlister, and SusanM. Broniarczyk (2005), “Perception of Assortment Variety:The Effects of Congruency between Consumers’ Internal andRetailers’ External Organization,” Journal of Retailing, 81(2), 159–69.

Nelson, Douglas L., Valier Reed, and John R. Walling (1976),“Pictorial Superiority Effect,” Journal of Experimental Psy-chology: Human Learning and Memory, 2 (5), 523–28.

Novemsky, Nathan, and Rebecca Ratner (2003), “The Time Courseand Impact of Consumers’ Erroneous Beliefs about HedonicContrast,” Journal of Consumer Research, 29 (March)507–16.

Paivio, Allan (1971), Imagery and Verbal Processes, New York:Holt, Rinehart, & Winston.

——— (1986), Mental Representations: A Dual-Coding Ap-proach, New York: Oxford University Press.

Preacher, Kristopher J., and Andrew F. Hayes (2004), “SPSS andSAS Procedures for Estimating Indirect Effects in SimpleMediation Models,” Behavior Research Methods, 36 (4),717–31.

Roediger, Henry, and Melissa Guynn (1996), “Retrieval Pro-cesses,” in Memory, ed. E. Bjork and R. Bjork, San Jose:California Academic Press, 197–236.

Ronis, David L., and J. Frank Yates (1987), “Components of Prob-ability Judgment Accuracy: Individual Consistency and Ef-

fects of Subject Matter and Assessment Method,” Organi-zational Behavior and Human Decision Processes, 40 (2),193–218.

Scheibehenne, Benjamin, Rainier Greifeneder, and Peter M. Todd(2010), “Can There Ever Be Too Many Options? A Meta-Analytic Review of Choice Overload,” Journal of ConsumerResearch, 3 (October), 409–25.

Schwartz, Barry, Andrew Ward, John Monterosso, Sonja Lyubom-irsky, Katherine White, and Donald R. Lehman (2002), “Max-imizing versus Satisficing: Happiness Is a Matter of Choice,”Journal of Personality and Social Psychology, 83 (5),1178–97.

Sharps, Matthew J., and Michael A. Nunes (2002), “Gestalt andFeature-Intensive Processing: Toward a Unified Model of Hu-man Information Processing,” Current Psychology: Devel-opmental, Learning, Personality, Social, 21 (1), 68–84.

Stenberg, George, Susanne Wiking, and Mats Dahl (1998), “Judg-ing Words at Face Value: Interference in a Word ProcessingTask Reveals Automatic Processing of Affective Facial Ex-pressions,” Cognition and Emotion, 12 (6), 755–82.

Tobii Technology (2010), “Tobii Studio Software User Manual,”September, http://www.tobii.com.

Trope, Yaacov, and Nira Liberman (2010), “Construal-Level The-ory of Psychological Distance,” Psychological Review, 117(2), 440–63.

Unnava, H. Rao, and Robert E. Burnkrant (1991), “An Imagery-Processing View of the Role of Pictures in Print Advertise-ments,” Journal of Marketing Research, 28 (2), 226–31.

Veryzer, Robert W., and J. Wesley Hutchinson (1998), “The In-fluence of Unity and Prototypicality on Aesthetic Responseto New Product Designs,” Journal of Consumer Research, 24(4), 374–85.

Young, Michael E., and Edward A. Wasserman (2001), “Entropyand Variability Discrimination,” Journal of Experimental Psy-chology: Learning, Memory, and Cognition, 27 (1), 278–93.

Zhao, Xinshu, John G. Lynch, and Qimei Chen (2010), “Recon-sidering Baron and Kenny: Myths and Truths about MediationAnalysis,” Journal of Consumer Research, 37 (2), 197–206.


http://www.tobii.com
